Not logged inRybka Chess Community Forum
Up Topic The Rybka Lounge / Computer Chess / Deep Fritz 7 vs. Houdini 1.5 x64 at ~1 hr per move 1/2-1/2
- - By turbojuice1122 (Gold) Date 2012-05-15 10:43
In line with Dragon49's match between Zappa Mexico II and Houdini, I recently played the following game with no opening books and no tablebases (tablebases would have made the draw quicker and helped Deep Fritz 7) at time controls of 15 hours per 30 moves plus 15 minutes per move.  I kept both engines loaded, gave each 1 GB of RAM, but had ponder off.  The computer is a Q6600 at 2.4 GHz.  Deep Fritz 7 made a nice late middlegame move, then blundered, but then was helped by a Houdini blunder to save the draw.

[Event "Microsoft, 1440'/30+1440'/30+1440'"]
[Site "Microsoft"]
[Date "2012.05.08"]
[Round "1"]
[White "Deep Fritz 7"]
[Black "Houdini 1.5a x64 4-CPU"]
[Result "1/2-1/2"]
[ECO "B34"]
[WhiteElo "3100"]
[BlackElo "3400"]
[Annotator "0.22;0.03"]
[PlyCount "155"]
[TimeControl "30/86400+900:30/86400+900:86400+900"]

1. e4 {0.22/21 5634} c5 {0.03/28 6428} 2. Nf3 {(Nc3) 0.38/20 4478} Nc6 {
0.03/28 6315} 3. d4 {(Bb5) 0.28/20 5144} cxd4 {0.04/28 3511} 4. Nxd4 {
0.28/20 4725} g6 {(d5) 0.09/28 7078} 5. Nc3 {(c4) 0.44/19 2960} Bg7 {
0.03/29 3346} 6. Be3 {0.25/19 4154} Nf6 {0.01/28 3168} 7. Be2 {
(f4) 0.28/19 3980} O-O {0.03/28 3057} 8. O-O {(f4) 0.22/19 5973} d5 $1 {
The obvious way to open up the position, but one that requires calculating all
of the tactics to the very end.  This would be quite difficult for a human,
but easy for a computer. -0.03/292903} 9. Nxd5 {(Bf3) 0.22/20 4503} Nxe4 {
(Nxd5) 0.00/30 3010} 10. Nxc6 {0.06/20 5997} bxc6 {-0.01/29 0} 11. Nb4 {
0.13/20 4178} Bxb2 {-0.01/30 3910} 12. Qxd8 {0.19/20 3715} Rxd8 {-0.03/29 5}
13. Rab1 {(Nxc6) 0.13/21 4837} Nc3 {-0.04/33 3467} 14. Bf3 {0.06/21 2550} Nxb1
{-0.04/31 26} 15. Rxb1 {0.13/20 671} Ba3 {-0.04/34 7361} 16. Nxc6 {0.09/21 1238
} Be6 {-0.05/33 3439} 17. Nxd8 {0.09/21 2882} Rxd8 {Now the tactics are over,
and black has a slight advantage due to having fewer pawn islands and the
ability to easily contain white's c-pawn with bishops and rook.  It should
only be enough for a draw, though. -0.06/31 2} 18. Be2 {(Rb5) 0.06/21 3730} Rc8
{-0.05/32 6358} 19. Bd3 {0.06/22 5383} Bd6 {-0.08/31 4687} 20. c4 {0.06/22 5310
} Bc5 {(a5) -0.10/33 7097} 21. Bxc5 {0.06/23 6197} Rxc5 {-0.11/31 17} 22. Rb3 {
0.06/22 4322} Ra5 {-0.08/34 7761} 23. a3 {0.09/23 2270} Kg7 {-0.09/33 5248} 24.
f4 {0.09/23 3759} Bd7 {(Kf6) -0.09/33 3462} 25. Kf2 {0.00/22 3611} Kf6 {
-0.11/33 6402} 26. g3 {(Be4) 0.00/21 1802} e5 {-0.13/32 2953} 27. Be2 $6 {
(Ke3) This is a strange move that makes things somewhat difficult for white.
Ke3 was expected and looks like the more straightforward path to hold equality.
-0.06/22 3471} exf4 {(Be6) -0.16/32 2921} 28. gxf4 {-0.09/23 2327} Ke7 {
-0.18/32 3182} 29. Ke3 {-0.06/21 908} Kd6 {-0.16/32 2732} 30. Rd3+ {
0.00/22 2218} Kc7 {-0.16/33 3556} 31. Rc3 {(h4) 0.00/24 4633} Kb6 {
(Ra6) -0.16/32 3432} 32. Kd4 {-0.06/24 3963} Rf5 {-0.16/33 3004} 33. c5+ $1 {
(Ke3) This move, combined with the next, create a very nice blockade that
should work to hold the draw with correct play.  The king cannot get past his
own third rank without a very long march. -0.12/235720} Kc6 {-0.24/32 3489} 34.
Bf3+ {0.00/24 3492} Kc7 {-0.24/30 2} 35. Ke3 {0.00/24 4135} Bc6 {-0.20/32 5962}
36. Bd1 {(Be2) 0.00/25 4067} Rd5 {(f6) -0.24/32 3175} 37. Be2 {0.00/24 3783}
Rd8 {(f6) -0.23/32 4186} 38. Bd3 {0.00/24 3616} Rd7 {(f6) -0.23/32 12184} 39.
Bc2 {(h4) 0.00/23 3207} Rd5 {(f6) -0.18/32 18866} 40. Bb3 {0.00/25 5308} Rh5 {
(Rf5) -0.18/32 2394} 41. Bxf7 {-0.19/24 7657} Rxh2 {-0.19/31 2183} 42. Rd3 {
(Be6) 0.00/23 5190} Rh3+ {-0.18/32 2246} 43. Kd4 {(Kd2) 0.00/24 2572} Rh4 {
-0.18/33 2723} 44. Ke3 {0.00/23 4527} Rh1 {(Rh3+) -0.14/31 2558} 45. Kd4 {
(Kd2) 0.00/21 3898} Bd7 {-0.09/31 2949} 46. Bg8 {(Re3) 0.00/22 2358} h5 {
(Rh4) -0.18/32 2318} 47. Bd5 {(Rg3) 0.00/23 4372} Rf1 {(Rg1) -0.19/33 4029} 48.
Ke5 $2 {(Rf3) The text move should have lost.  After Rf3, black's most
straightforward route is to acquiese to the exchange of rooks with a dead
drawn ending.  Interestingly, Fritz 5.32 finds the correct move and is easily
able to hold a draw against Houdini at a faster time control on a single-core
computer. 0.00/22 2680} h4 {(Rc1) -0.27/33 2229} 49. Rb3 {(Bg2) -0.31/23 6416}
Bc8 {-0.30/33 2623} 50. Rb2 {-0.31/23 4549} h3 {-0.24/33 2984} 51. Be4 {
-0.37/22 2284} Re1 {-0.23/34 6511} 52. Kd4 {-0.47/23 3509} Rd1+ $2 {(Rg1) Now b
lack returns the favor.  After ...Rg1!, black wins after a bit of maneuvering
(at least, Deep Fritz 7 is never able to hold the draw, and I cannot find a
path to do so). -0.21/32 2955} 53. Ke5 {-0.25/22 2602} Rg1 {-0.18/33 2749} 54.
Rh2 {-0.25/22 2202} a5 {-0.14/32 2583} 55. a4 {-0.22/24 4738} Bd7 {
(Re1) -0.11/33 2034} 56. Kf6 {0.00/24 3398} Rg4 {(Rf1) -0.09/34 2512} 57. Ke5 {
0.00/25 4573} Rg3 {(Rg1) -0.04/35 2688} 58. c6 {0.00/24 1722} Bg4 {
(Bc8) -0.03/35 2346} 59. Rb2 {0.00/23 860} Bc8 {-0.02/35 2398} 60. Rh2 {
0.00/23 959} Rg1 {(Bg4) 0.01/35 2749} 61. Kf6 {0.00/25 4487} Rf1 {
(Rg4) 0.01/35 3527} 62. Kg5 {0.00/25 3845} Rc1 {(Rg1+) 0.01/34 3364} 63. Bxg6 {
0.00/24 3320} Rg1+ {0.01/35 2034} 64. Kf6 {0.00/25 3213} Rg2 {0.01/35 2680} 65.
Rh1 {0.00/25 3407} Rf2 {(h2) 0.01/35 2600} 66. Ke5 {(f5) -0.06/23 2233} h2 {
0.01/34 3626} 67. Be4 {(f5) 0.00/23 2625} Bh3 {0.02/37 5069} 68. Kd4 {
(Bd5) 0.00/23 2188} Ra2 {0.01/35 1879} 69. Kc3 {(Kc5) 0.00/22 2688} Ra3+ {
(Re2) 0.00/35 1901} 70. Kb2 {-0.12/24 4185} Rxa4 {0.00/37 2430} 71. Bd5 {
(Rxh2) -0.09/22 2113} Rd4 {(Bf5) 0.00/36 1852} 72. Rxh2 {-0.12/23 3710} Rxd5 {
(Bf5) 0.00/38 2179} 73. Rxh3 {-0.03/24 896} Kxc6 {(Rd2+) 0.00/43 1768} 74. Rf3
{(Rh6+) 0.09/25 1733} Rd2+ {(Rf5) 0.00/41 2042} 75. Ka3 {(Kb3) 0.16/26 2532}
Kd7 {(Kd5) 0.00/46 1752} 76. f5 {(Rc3) 0.13/26 1744} Ke7 {0.00/50 1840} 77. f6+
{0.06/25 1627} Kf7 {0.00/51 2367} 78. Ka4 {adjud. Draw accepted 0.09/26 2897}
1/2-1/2
Parent - - By turbojuice1122 (Gold) Date 2012-05-15 11:00
I should mention that according to the CEGT 40/4 list, these engines are 500 elo points apart.  Even at 40 moves in 2 hours, Deep Rybka 4.1 is over 300 points ahead of Deep Fritz 8 on the SSDF rating list, which would equate to nearly 400 points for Houdini 1.5 compared with Deep Fritz 7.
Parent - - By RFK (Gold) Date 2012-05-16 16:35
There has to be some kind of logical reason for this discrepancy.
Parent - - By turbojuice1122 (Gold) Date 2012-05-16 20:46
When you get to differences of many hundreds of Elo points in computer chess, there are going to be large uncertainties involved.  Also, the 500 points was at a short time control, and the 400 points is at a longer time control.  My belief is that at the time controls I'm dealing with here, we're talking about an Elo difference of 100-150 points or so due to approaching the draw horizon.
Parent - By Uri Blass (*****) Date 2012-05-18 11:05
I disagree.

I believe that the difference is smaller with longer time control but I expect 300 elo difference even at 1 hour per move.
Parent - - By Uly (Gold) Date 2012-05-15 13:35
Wow, very interesting result. I also think it's interesting to think this is one minute/move from the future, and to think about which time control would be needed for top engine of today to play perfect chess in 90% of their games.
Parent - By turbojuice1122 (Gold) Date 2012-05-15 20:25
It is one minute per move from the future, though not as close as this type of time control would have been 15 years ago.  Notice that the speed of the desktops bought at a typical electronics store has not increased significantly over the past 4 years, while in the 4 years before that we went from single-core to quad-core.
Parent - - By dragon49 (****) Date 2012-05-15 16:13
I'm glad I'm influencing others!
Parent - By turbojuice1122 (Gold) Date 2012-05-15 20:26
Indeed. :smile:
Parent - - By Uri Blass (*****) Date 2012-05-18 11:16
I wonder if houdini has a bug in games relative to analysis because at analysis mode even 1 cpu houdini likes 52...Rg1 that you mention as a winning move(I did not try to get to depth 32 and I wonder if houdini changes its mind at high depth to the game move 52...Rd1+).

Note that I fully expect houdini32 bits 1 cpu at 3 minutes per move to be stronger than Deep Fritz7 at 1 hour per move.

Deep Fritz 7 - Houdini 1.5a x64 4-CPU, Microsoft, 1440'/30+1440'/30+1440' 2012
2b5/p1k5/6p1/2P5/3KBP2/P6p/1R6/4r3 b - - 0 1


Analysis by Houdini 1.5 w32:

52...Re1-d1+ 53.Kd4-e5 Rd1-c1 54.Be4xg6 Rc1xc5+ 55.Ke5-f6
  ³  (-0.40)   Depth: 4/12   00:00:00
52...Bc8-f5 53.Be4-f3 Re1-b1 54.Rb2xb1 Bf5xb1
  ³  (-0.45)   Depth: 4/15   00:00:00
52...Bc8-f5 53.Be4-f3 Re1-b1 54.Rb2-e2 Rb1-d1+ 55.Kd4-e5
  ³  (-0.60)   Depth: 5/15   00:00:00
52...Bc8-f5 53.Be4xf5 g6xf5 54.Rb2-h2 Re1-e4+ 55.Kd4-d5 Re4-e3
  ³  (-0.29)   Depth: 6/15   00:00:00  4kN
52...Bc8-f5 53.Be4xf5 g6xf5 54.Rb2-h2 Re1-e4+ 55.Kd4-d5 Re4-e3 56.a3-a4
  ³  (-0.28)   Depth: 7/15   00:00:00  5kN
52...Re1-d1+
  ³  (-0.33)   Depth: 7/15   00:00:00  6kN
52...Re1-d1+ 53.Kd4-e3 Rd1-c1 54.Be4xg6 Rc1xc5 55.Ke3-e4 Kc7-d6 56.Rb2-d2+ Kd6-e6
  ³  (-0.30)   Depth: 7/18   00:00:00  10kN
52...Re1-g1 53.Kd4-e5 Rg1-g3 54.a3-a4 Rg3-g1 55.Rb2-d2 Rg1-e1
  ³  (-0.32)   Depth: 7/18   00:00:00  12kN
52...Re1-g1 53.Kd4-e5 Rg1-g3 54.a3-a4 Rg3-g1 55.f4-f5 g6xf5 56.Be4xf5 Bc8xf5 57.Ke5xf5
  ³  (-0.33)   Depth: 8/18   00:00:00  13kN
52...Re1-g1 53.Kd4-e5 a7-a5 54.Rb2-d2 Rg1-g3 55.a3-a4 Bc8-d7 56.Rd2-b2 Bd7-c6 57.Rb2-b6
  ³  (-0.38)   Depth: 9/22   00:00:00  33kN
52...Re1-g1 53.Kd4-e5
  ³  (-0.33)   Depth: 10/22   00:00:00  41kN
52...Re1-g1 53.Kd4-e5 a7-a5 54.Rb2-h2 Rg1-e1 55.Ke5-d4 Re1-d1+ 56.Kd4-e5 Rd1-g1 57.f4-f5 g6xf5 58.Be4xf5 Bc8xf5 59.Ke5xf5
  ³  (-0.32)   Depth: 10/22   00:00:00  61kN
52...Re1-g1 53.Kd4-e5 a7-a5 54.Rb2-h2 Rg1-e1 55.Ke5-d4 Re1-d1+ 56.Kd4-e5 Rd1-g1 57.f4-f5 g6xf5 58.Be4xf5 Bc8xf5 59.Ke5xf5
  ³  (-0.32)   Depth: 11/22   00:00:00  74kN
52...Re1-g1
  ³  (-0.37)   Depth: 12/27   00:00:01  159kN
52...Re1-g1 53.Rb2-h2 Bc8-d7 54.Kd4-e5 Rg1-e1 55.Ke5-d4 Re1-d1+ 56.Kd4-e5 Rd1-g1 57.a3-a4 Rg1-e1 58.Ke5-d4 Re1-d1+ 59.Kd4-e5 Rd1-g1 60.a4-a5 Rg1-c1
  ³  (-0.37)   Depth: 12/27   00:00:01  192kN
52...Re1-g1 53.Rb2-h2
  ³  (-0.32)   Depth: 13/27   00:00:01  273kN
52...Re1-g1 53.a3-a4 Rg1-f1 54.Kd4-e5 Rf1-e1 55.Ke5-d4 Re1-g1 56.Kd4-e5 a7-a5 57.Ke5-f6 Rg1-f1 58.Kf6-g5 Rf1-a1
  ³  (-0.34)   Depth: 13/27   00:00:01  460kN
52...Re1-g1
  ³  (-0.39)   Depth: 14/27   00:00:01  642kN
52...Re1-g1 53.a3-a4 Rg1-d1+ 54.Kd4-e5 Rd1-e1 55.Ke5-d4 Re1-g1 56.Kd4-e5 a7-a6 57.Rb2-d2 a6-a5 58.Rd2-e2 Rg1-c1 59.Be4xg6 Rc1xc5+ 60.Ke5-f6 Rc5-c6+ 61.Kf6-g5
  ³  (-0.33)   Depth: 14/27   00:00:01  737kN
52...Re1-g1 53.a3-a4 Rg1-g3 54.Kd4-e5 a7-a5 55.Rb2-h2 Rg3-g1 56.Ke5-f6 Rg1-f1 57.Kf6-e5 Rf1-e1 58.Ke5-d4 Re1-g1 59.Kd4-e3 Bc8-d7 60.Ke3-f2 Rg1-a1 61.Be4xg6 Ra1-a2+ 62.Kf2-g1 Ra2xa4 63.f4-f5 Ra4-a1+ 64.Kg1-f2
  ³  (-0.33)   Depth: 15/30   00:00:02  1076kN
52...Re1-g1 53.a3-a4 Rg1-g3 54.Kd4-e5 a7-a5 55.Rb2-h2 Rg3-g1 56.c5-c6 Rg1-e1 57.Ke5-d4 Re1-d1+ 58.Kd4-e5 Rd1-d6 59.f4-f5 g6xf5 60.Be4xf5 Bc8xf5 61.Ke5xf5 Rd6-d3 62.Kf5-e4
  ³  (-0.31)   Depth: 16/30   00:00:02  1422kN
52...Re1-g1 53.a3-a4 Rg1-g3 54.Kd4-e5 a7-a5 55.Rb2-h2 Rg3-g1 56.c5-c6 Rg1-e1 57.Ke5-d4 Re1-d1+ 58.Kd4-e5 Rd1-g1 59.Ke5-f6 Rg1-f1 60.Kf6-g5 Rf1-a1 61.Be4xg6 Ra1-g1+ 62.Kg5-h5 Bc8-g4+ 63.Kh5-h4
  ³  (-0.29)   Depth: 17/37   00:00:03  2479kN
52...Re1-g1 53.a3-a4 Rg1-f1 54.Kd4-e3 Rf1-e1+ 55.Ke3-d3 Re1-c1 56.Be4xg6 Rc1xc5 57.Rb2-h2 Rc5-a5 58.f4-f5 Ra5xa4 59.Rh2-c2+ Kc7-d7 60.f5-f6 Bc8-a6+ 61.Kd3-e3 Kd7-e6 62.f6-f7 Ke6-e7 63.Rc2-c7+ Ke7-f8 64.Rc7xa7 Ba6-b5
  ³  (-0.29)   Depth: 18/42   00:00:09  5085kN
52...Re1-g1 53.a3-a4 Rg1-f1 54.Kd4-e3 Rf1-e1+ 55.Ke3-d3 Re1-c1 56.Be4xg6 Rc1xc5 57.Rb2-h2 Rc5-a5 58.f4-f5 Ra5xa4 59.Rh2-c2+ Kc7-d7 60.f5-f6 Bc8-b7 61.Kd3-e3 Ra4-b4 62.Rc2-e2 Bb7-d5 63.Re2-h2 Bd5-g2 64.Bg6-f5+ Kd7-e8
  ³  (-0.29)   Depth: 19/44   00:00:15  8366kN
52...Re1-g1 53.a3-a4 Rg1-f1 54.Kd4-e3 Rf1-e1+ 55.Ke3-d4 Re1-g1 56.Rb2-h2 Rg1-g3 57.Kd4-e5 Bc8-b7 58.c5-c6 Bb7-c8 59.Ke5-f6 a7-a5 60.Kf6-e5 Rg3-g1 61.Ke5-f6 Rg1-g4 62.Kf6-e5 Rg4-g3 63.Rh2-h1 Bc8-g4 64.Ke5-f6 Bg4-e2 65.Rh1-h2 Be2-d1 66.f4-f5 Rg3-e3 67.f5xg6 Re3xe4 68.g6-g7
  =  (-0.25)   Depth: 20/44   00:00:22  12415kN
52...Re1-g1
  ³  (-0.30)   Depth: 21/44   00:00:34  22937kN
52...Re1-g1 53.Rb2-h2 Rg1-g3 54.a3-a4 Bc8-b7 55.c5-c6 Bb7-a6 56.a4-a5 Ba6-c8 57.Kd4-e5 Rg3-g1 58.Ke5-f6 Rg1-f1 59.Kf6-e5 Rf1-e1 60.Ke5-d4 Re1-d1+ 61.Kd4-e3 Rd1-g1 62.Ke3-f2 Rg1-a1 63.Be4xg6 Ra1xa5 64.Bg6-e4 Ra5-a4 65.Kf2-e3 Ra4-a3+ 66.Ke3-d4 a7-a5 67.f4-f5
  ³  (-0.26)   Depth: 21/53   00:00:45  32496kN
52...Re1-g1 53.Rb2-h2 Rg1-g3 54.a3-a4 Bc8-b7 55.c5-c6 Bb7-a6 56.a4-a5 Ba6-c8 57.Kd4-e5 Rg3-g1 58.Ke5-f6 Rg1-f1 59.Kf6-e5 Rf1-e1 60.Ke5-d4 Re1-d1+ 61.Kd4-e3 Rd1-g1 62.Ke3-f2 Rg1-a1 63.Be4xg6 Ra1xa5 64.Bg6-e4 Ra5-a4 65.Kf2-e3 Ra4-a3+ 66.Ke3-d4 a7-a5 67.f4-f5
  ³  (-0.26)   Depth: 22/53   00:01:00  41805kN
52...Re1-g1 53.Rb2-h2 Rg1-g3 54.a3-a4 a7-a6 55.Kd4-e5 a6-a5 56.Ke5-f6 Rg3-g4 57.Kf6-e5 Rg4-g1 58.Ke5-d4 Bc8-d7 59.Kd4-e5 Rg1-e1 60.Ke5-d4 Re1-f1 61.Kd4-e3 Rf1-g1 62.Ke3-f3 Rg1-a1 63.Be4xg6 Ra1-a3+ 64.Kf3-e4 Ra3xa4+ 65.Ke4-e3 Ra4-a3+ 66.Ke3-d4 Ra3-f3 67.Kd4-e5 Rf3-g3 68.Bg6-e4
  ³  (-0.26)   Depth: 23/64   00:01:40  69727kN
52...Re1-g1
  ³  (-0.30)   Depth: 24/64   00:02:56  134730kN
52...Re1-g1 53.a3-a4 Rg1-a1 54.Kd4-e5 Ra1xa4 55.Be4xg6 Ra4-c4 56.f4-f5 Rc4xc5+ 57.Ke5-d4 Kc7-d6 58.Rb2-f2 a7-a5 59.f5-f6 Bc8-e6 60.f6-f7 Rc5-c4+ 61.Kd4-d3 Be6xf7 62.Bg6xf7 Rc4-h4 63.Rf2-h2 a5-a4 64.Kd3-c2 Kd6-e5 65.Rh2xh3 Rh4xh3 66.Kc2-b2 Rh3-e3 67.Kb2-a2 Re3-e2+ 68.Ka2-a1
  ³  (-0.33)   Depth: 24/64   00:07:18  353337kN

(,  18.05.2012)
Parent - By Fulcrum2000 (****) Date 2012-05-18 12:35

> I wonder if houdini has a bug in games relative to analysis because at analysis mode even 1 cpu houdini likes 52...Rg1 that you mention as a winning move


Default game mode has contempt 1, default analysis has contempt 0.
Maybe that's (part) of the reason?.
Parent - By turbojuice1122 (Gold) Date 2012-05-18 21:16
Houdini liked it on my computer during the game, too...until it "thought for too long". :smile:
Parent - - By siah (***) Date 2012-05-19 06:09
Can you tell the blunders of both engines?
Parent - By turbojuice1122 (Gold) Date 2012-05-19 10:49
White's 27th move (Deep Fritz 7) was a weak move, white's 48th move (Deep Fritz 7) was a blunder that should have lost, and black's 52nd move (Houdini 1.5) was a blunder that made the position drawn again.
- - By turbojuice1122 (Gold) Date 2012-05-16 11:38
I have started the second game, this time at 24 hrs per 20 moves with 15 minutes per move, repeated.  This time, Deep Fritz 7 has black, and we have an interesting D44 Semi-Slav.  This will be a great game, so I've included the first few moves:

1. d4 Nf6 2. c4 e6 3. Nf3 d5 4. Bg5 dxc4 5. e3 b5 6. a4 c6 7. Nc3 Bb4 8. g3

These programs are playing the opening like World Champions. :smile:  Also, interestingly, Houdini's score has been going down, and is near zero; Deep Fritz 7 thinks that it has an advantage with black.  I recall that theory says black has at least equal chances here.
Parent - - By Uly (Gold) Date 2012-05-16 12:53
Predicting that in one of these games, Houdini will win because Fritz will think he's winning on a draw position and overreach, let's see :smile:
Parent - By turbojuice1122 (Gold) Date 2012-05-16 16:11
It might not be that so much as the fact that this opening can get very nasty quickly--it's amazing that the programs have played this from the starting position.  In such nasty situations, the better program is going to have an advantage, but we'll have to see whether that's relevant at this level of play.
- - By turbojuice1122 (Gold) Date 2012-05-16 23:36
The game with Deep Fritz 7 having black continues:

1. d4 Nf6 2. c4 e6 3. Nf3 d5 4. Bg5 dxc4 5. e3 b5 6. a4 c6 7. Nc3 Bb4 8. g3 Bb7 9. Bg2 Nbd7 10. 0-0 a6 11. Ne4 h6 12. Nxf6+ gxf6!?

Deep Fritz 7 has played the opening very well for an unaided engine with no book.  Even Houdini agrees that black has a small advantage here.  Deep Fritz 7 says =/+.  I fear that black might be overreaching with that last recapture, though...
Parent - - By turbojuice1122 (Gold) Date 2012-05-17 22:24
The game continues, and is quite exciting:

1. d4 Nf6 2. c4 e6 3. Nf3 d5 4. Bg5 dxc4 5. e3 b5 6. a4 c6 7. Nc3 Bb4 8. g3 Bb7 9. Bg2 Nbd7 10. 0-0 a6 11. Ne4 h6 12. Nxf6+ gxf6!? 13. Bh4 Qc7!? 14. Nd2 f5!? 15. e4 0-0 16. exf5 exf5 17. Bh3 Qd6 18. Bxf5 Nf6 19. Ne4 Nxe4 20. Bxe4 Qe6 21. Bg2 with a score of -0.03 according to Deep Fritz 7 and 0.25 (and dropping toward zero over the last few moves) according to Houdini.  After 14...f5!?, Deep Fritz 7 scored the position -0.47.  The question is whether it can actually achieve anything with this queenside attack.  I think that Houdini's kingside chances aren't very high, even though black's kingside doesn't look too good.
Parent - - By RFK (Gold) Date 2012-05-18 04:04
There has to be some way to copy and paste this...
Parent - - By Pia (****) Date 2012-05-18 05:03
Oui, if Fritz 11 GUI, start new game, then paste game.

But commentary missing afterwards:

[Event "?"]
[Site "?"]
[Date "????.??.??"]
[Round "?"]
[White "New game"]
[Black "?"]
[Result "*"]
[PlyCount "41"]

1. d4 Nf6 2. c4 e6 3. Nf3 d5 4. Bg5 dxc4 5. e3 b5 6. a4 c6 7. Nc3 Bb4 8. g3 Bb7
9. Bg2 Nbd7 10. O-O a6 11. Ne4 h6 12. Nxf6+ gxf6 13. Bh4 Qc7 14. Nd2 f5 15. e4
O-O 16. exf5 exf5 17. Bh3 Qd6 18. Bxf5 Nf6 19. Ne4 Nxe4 20. Bxe4 Qe6 21. Bg2 *
Parent - By RFK (Gold) Date 2012-05-18 05:34
Thank you!
Parent - - By Uly (Gold) Date 2012-05-19 13:01

> 0.25 (and dropping toward zero over the last few moves) according to Houdini.  After 14...f5!?, Deep Fritz 7 scored the position -0.47.


Well, I'm going to claim right prediction here :smile: (though I guess I expected a more more dramatic score)

I'm still not impressed by what the engine's are showing here (strengthwise), I think a strong centaur at 1 hour for the entire game TC could beat either entity.
Parent - - By turbojuice1122 (Gold) Date 2012-05-19 15:05
Are you saying that a strong centaur at one hour per game would beat both Deep Fritz 7 and Houdini 1.5 at one hour per game, or that a strong centaur at one hour per game would beat Deep Fritz 7 and Houdini 1.5 at 20 moves in 24 hours?
Parent - - By Uly (Gold) Date 2012-05-20 07:29
The latter (with the centaur getting time odds), though I guess it would get complicated as of course the centaur could ponder while it's the engine's time to move...
Parent - By turbojuice1122 (Gold) Date 2012-05-20 13:14
If the centaur is not allowed to ponder, he would be very lucky to draw Deep Fritz 7 at that time control, even if he gets the white pieces.
Parent - - By turbojuice1122 (Gold) Date 2012-05-19 15:07

> Well, I'm going to claim right prediction here


Yes, your prediction was quite good, though we'll have to see whether Houdini is able to win this.  I think that both engines' scores are too generous for white--black can probably hold for quite awhile by shuffling rooks back and forth, with the occasional necessary bishop or king move.
Parent - - By Uly (Gold) Date 2012-05-20 07:33

> we'll have to see whether Houdini is able to win this.


This is very interesting, as I've read people claiming that Houdini's evaluation is super-accurate when it comes to winning and they hadn't seen it giving a +2.00 eval to a position it couldn't win, this would be a counter-example.
Parent - By turbojuice1122 (Gold) Date 2012-05-20 13:15
Houdini will win this, though I feel like somewhere, Deep Fritz 7 went wrong with a bishop shuffle.
Parent - - By turbojuice1122 (Gold) Date 2012-05-19 10:55
It looks like black's risky play has backfired, and white will probably win this game.  Not a dominating performance (more like the type of game I would expect for 50 or 100 elo difference in engines), but a game full of fireworks in which the stronger player will most likely come out on top.

1. d4 Nf6 2. c4 e6 3. Nf3 d5 4. Bg5 dxc4 5. e3 b5 6. a4 c6 7. Nc3 Bb4 8. g3 Bb7 9. Bg2 Nbd7 10. O-O a6 11. Ne4 h6 12. Nxf6+ gxf6 13. Bh4 Qc7 14. Nd2 f5 15. e4 O-O 16. exf5 exf5 17. Bh3 Qd6 18. Bxf5 Nf6 19. Ne4 Nxe4 20. Bxe4 Qe6 21. Bg2 Rab8?! 22. Qf3 f5?! 23. Rfd1 Kh7 24. Qf4 Bd6 25. Qd2 Qf7 26. d5! cxd5 27. axb5 axb5 28. Ra7 Bc5 29. Bxd5 Qg7 30. Re1 Bxa7?! 31. Re7 Bxd5 32. Rxg7+

Deep Fritz 7 scores this as +0.91, while Houdini 1.5 scores this as +2.64.  A good and exciting game, though.
Parent - By turbojuice1122 (Gold) Date 2012-05-20 13:18
It looks like the second game will end soon with a win for white (Houdini).  After the opening, this quickly went into the late middlegame, where Houdini/Rybka are going to excel.  There is no doubt that Deep Fritz 7 played the opening MUCH better than Houdini, giving black an advantage that I think might have been enough to win if I had switched to Rybka in the late middlegame (i.e. before move 20 in this game).  It was a very interesting game, though, and I will post it when it is officially complete.
- By turbojuice1122 (Gold) Date 2012-05-20 19:37
I have adjudicated the second game, as both engines agree that black will have to give up a rook to stop the b-pawn.  Here is the game:

[Event "Microsoft, 1440'/20+0'/0+0'0"]
[Site "Microsoft"]
[Date "2012.05.15"]
[Round "1"]
[White "Houdini 1.5a x64 4-CPU"]
[Black "Deep Fritz 7"]
[Result "1-0"]
[ECO "D44"]
[WhiteElo "3400"]
[BlackElo "3100"]
[Annotator "0.22;0.12"]
[PlyCount "95"]
[TimeControl "20/86400+900:0/0:0/0"]

1. d4 {0.22/28 9035} Nf6 {0.12/21 5634} 2. c4 {(Nf3) 0.25/28 3914} e6 {
0.09/20 4086} 3. Nf3 {0.23/28 6800} d5 {(Bb4+) 0.09/20 7513} 4. Bg5 {
(Nc3) 0.27/27 5090} dxc4 {0.03/19 3651} 5. e3 {(e4) 0.25/28 3915} b5 {
(c5) -0.06/19 5778} 6. a4 {0.18/28 5868} c6 {-0.03/19 7135} 7. Nc3 {
(Be2) 0.18/27 5060} Bb4 {(Qb6) -0.16/19 7676} 8. g3 {(Be2) 0.11/26 4138} Bb7 {
-0.31/19 4833} 9. Bg2 {0.05/28 7330} Nbd7 {(a6) -0.28/19 4778} 10. O-O {
-0.01/25 3718} a6 {-0.31/18 6368} 11. Ne4 {(Ne5) -0.11/25 5606} h6 {
(0-0) -0.44/18 3503} 12. Nxf6+ {(Bxf6) -0.08/27 3781} gxf6 {
(Nxf6) -0.41/19 2272} 13. Bh4 {0.05/28 4228} Qc7 {(c5) -0.34/19 10290} 14. Nd2
{0.12/26 6579} f5 {(Rg8) -0.47/18 4777} 15. e4 {0.37/28 3913} O-O {
-0.28/17 4439} 16. exf5 {0.37/26 40} exf5 {0.09/19 7055} 17. Bh3 {0.25/30 13492
} Qd6 {0.09/19 6194} 18. Bxf5 {0.25/28 46} Nf6 {0.12/18 3063} 19. Ne4 {
0.32/28 7135} Nxe4 {0.00/20 2891} 20. Bxe4 {(Qg4+) 0.32/27 134} Qe6 {
-0.06/18 1315} 21. Bg2 {0.23/30 5683} Rab8 {(Qf5) 0.03/20 8156} 22. Qf3 {
0.30/28 8817} f5 {(Kg7) 0.03/21 22028} 23. Rfd1 {0.24/28 4989} Kh7 {
(a5) -0.03/19 7003} 24. Qf4 {0.37/29 4215} Bd6 {(a5) 0.09/20 3872} 25. Qd2 {
0.57/30 6912} Qf7 {0.09/20 6099} 26. d5 {(Re1) 0.54/29 12088} cxd5 {
0.00/20 3549} 27. axb5 {0.76/29 5971} axb5 {0.00/20 2988} 28. Ra7 {1.07/30 7500
} Bc5 {(Qg7) 0.41/21 9021} 29. Bxd5 {1.19/29 4396} Qg7 {0.37/20 2700} 30. Re1 {
1.19/27 6} Bxa7 {(f4) 0.53/20 2774} 31. Re7 {2.43/27 796} Bxd5 {0.91/22 5913}
32. Rxg7+ {2.64/32 4338} Kxg7 {1.47/22 5917} 33. Qxd5 {2.38/31 1} Rf7 {
1.41/21 5196} 34. Qe5+ {(Bd8) 2.81/32 12870} Kg6 {(Kg8) 1.41/22 6777} 35. Be7 {
3.17/30 5045} Rb7 {1.41/20 2356} 36. Bb4 {3.17/28 2} Bb6 {(Rbd7) 1.62/19 2371}
37. Qh8 {3.94/31 14153} Rbd7 {(f4) 2.22/19 4768} 38. Bc3 {5.00/27 1122} Kh5 {
2.50/18 849} 39. Qe8 {5.53/30 5259} b4 {(Bd8) 2.81/18 886} 40. Bxb4 {
6.44/26 1497} Bd4 {3.75/18 951} 41. Qe6 {6.66/25 457} Rb7 {4.34/20 9092} 42.
g4+ {(Kg2) 6.65/24 124} fxg4 {(Kxg4) 4.47/19 5739} 43. Qd5+ {6.89/23 83} Kg6 {
5.00/21 11091} 44. Qxd4 {6.91/22 41} Rf5 {(Kh7) 5.56/20 5203} 45. Qxc4 {
7.25/21 35} Rbf7 {5.91/20 3707} 46. Bc5 {7.39/20 3} Kh5 {(Kf6) 6.09/20 4091}
47. Bd4 {(b4) 7.49/19 2} Re7 {7.84/20 6270} 48. b4 {adjud. 7.49/18 0} 1-0
Up Topic The Rybka Lounge / Computer Chess / Deep Fritz 7 vs. Houdini 1.5 x64 at ~1 hr per move 1/2-1/2

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill